Visualised Vibrations

Registration of tremors using video analysis

Master Thesis (2026)
Author(s)

M. Thakur (TU Delft - Electrical Engineering, Mathematics and Computer Science)

Contributor(s)

J.C. van Gemert – Mentor (TU Delft - Electrical Engineering, Mathematics and Computer Science)

Robert van den Berg – Mentor (Erasmus MC)

M. Weinmann – Graduation committee member (TU Delft - Electrical Engineering, Mathematics and Computer Science)

Faculty
Electrical Engineering, Mathematics and Computer Science
More Info
expand_more
Publication Year
2026
Language
English
Graduation Date
26-11-2026
Awarding Institution
Delft University of Technology
Project
DSAIT 5000
Programme
Computer Science, Data Science and Artificial Intelligence Technology
Faculty
Electrical Engineering, Mathematics and Computer Science
Page Views
8
Reuse Rights

Other than for strictly personal use, it is not permitted to download, forward or distribute the text or part of it, without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license such as Creative Commons.

Abstract

Quantifying hand tremor from ordinary RGB videos offers a potentially accessible and contactless alternative to characterize patients with Chronic Inflammatory Demyelinating Polyradiculoneuropathy (CIDP). However, modern hand pose estimation methods are not optimized for the preservation of small, periodic movements. We investigate the reliability of monocular video-based tremor frequency estimation using a model-agnostic pipeline that extracts periodic motion from reconstructed hand trajectories. We compare four methods (MediaPipe, HAMBA, WiLoR, and STRIDE) on controlled recordings of emulated finger and wrist tremors, followed by more complex drawing activities and clinical recordings of CIDP patients. Under controlled conditions, WiLoR achieves the lowest IQR MAE of 0.12 Hz, while temporal refinement using STRIDE increases the error to 0.27 Hz, demonstrating that smoother hand reconstructions do not necessarily yield more accurate frequency estimates. Lastly, we investigate the reliability of tremor frequency estimates, and whether they can be assessed quantitatively without annotated data. We propose and apply metrics to controlled and clinical recordings, trying to characterize reconstruction reliability. The findings demonstrate the feasibility of our method under controlled conditions while highlighting the need for further research before extending to clinical settings.

Files

License info not available